Less than 3ms latency at 44.1kHz? Why...

[CreamWare4Ever]

Hello,

I'm fully satisfied with the power my Scope platforms deliver. I can load large number of VSTi:s without to compromise ASIO stability and they all play without clicks an pops. However, I want to try to get 1 ro 2 ms latency - but at 44,1kHz. Actually, Cubase calculate latency 2x so if it said 3ms in its ASIO panel - that means 6 ms total after I struck the keyboard.

Is it possible to override buffer by changing the Scope v.5.1.x drive? Of course, I'm not sure if such 'hack' is legal, only if it is then...

Any idea?

[garyb]

nope, that's it.
any less would be unstable.
there has to be time lag for the various processes that have to run. even the fastest system is less than realtime.

on the card, calculations are one pointed and almost instant. from ad/da to dsp and back to the real world is so fast as to be undectable by the ear(maybe). but once the computer is in the picture, there are numerous pieces of software which must communicate with several pieces of hardware, as well as each other, immediately, and only the cpu can do the calculations to make that possible. in order to get steady streaming audio with no hiccups, there's a minimum delay and unforunately, it's about 3ms(i use 4ms).

still, 6ms is the real world equivalent to being 6 feet from your speaker cabinet(1.8m), since sound travels about a foot per millisecond(.3m/millisecond). that's pretty usable. it shouldn't be the difference between a good and a great performance, nor a good and a bad one. (ok, that's my opinion..)

[CreamWare4Ever]

still, 6ms is the real world equivalent to being 6 feet from your speaker cabinet(1.8m), since sound travels about a foot per second(.3m/second

Sure, but in DAW world that is the double: input AND output latency. So when we are speaking about 3ms that is actually 6ms: I hit the key on keyboard >> 3ms delay BEFORE Cubase (or whatever) tells its VSTi to play; then ANOTHER 3ms for DAW to process output signal to speakers; than another 'delay' for sound to travel from speakers to your ears (not so important).

So my formula to calculate latency is:

Buffer / Frequency x 2 + SFSTE

NOTE: 'XSTE' variable is undefined as 'Sound from Speakers to Ears'
Example:

(128/44,1) x 2 + SFSTE = 5,8ms + SFSTE

Sometimes I found my self switching to 1ms at 96kHz. I personally can 'feel' (rather than hear) difference between 12ms >> 6ms >> 2ms absolutely. But under 6ms it doesn't matter. Still I know that even hardware have some latency. I think that I saw some software that can measure that but I'm not sure...

[astroman]

...However, I want to try to get 1 ro 2 ms latency - but at 44,1kHz. Actually, Cubase calculate latency 2x so if it said 3ms in its ASIO panel - that means 6 ms total after I struck the keyboard. ...

so you HIT the keyboard... any idea WHEN this happens ?
Midi has a very low clock rate of 31250 bit/sec which results in a theoretical capacity of roughly 1k notes/sec (4x8bits).
Even under perfect conditions any note could be 1ms early or late, but (imho) an average drift of 3 milliseconds is more likely for an unprocessed midi datastream, so there's few sense to reduce audio latency below that value.
That may explain why you don't notice much difference below 6 ms: midi data can be off up to 5 ms in real world gear.

cheers, Tom

[tgstgs]

this is a cubasevsti hac_k not a scope one;

good vibes

[JubaSky]

Gary,

Last time I've checked, speed of sound in air is about 340 metres per second...


Marcos.

[tgstgs]

was interested myself so i made a speedtest:

the midiroundtriplatency for a scopePCIboard on a win xp prof. 1.8 GHz dualcore is

848 microsecunds = 0.848ms

this is a averaged value taken from 100000 midievents send from and to host;

so you should have half is about 0.5ms for 1 way;

the audio back is dependant on ULLI settings but only once NOT twice!!

if your vsti takes longer -> contact its developer_

good vibes from vienna

[CreamWare4Ever]

was interested myself so i made a speedtest:

the midiroundtriplatency for a scopePCIboard on a win xp prof. 1.8 GHz dualcore is

848 microsecunds = 0.848ms

this is a averaged value taken from 100000 midievents send from and to host;

so you should have half is about 0.5ms for 1 way;

the audio back is dependant on ULLI settings but only once NOT twice!!

if your vsti takes longer -> contact its developer_

good vibes from vienna

Maybe we misunderstood each other!?

I'm talking about following:



You can see that it said '2.95ms' as Output latency but for VSTi input latency need to be calculated also. So in real world the ASIO latency of '2.95ms' is at the end '5,9ms'.

Cheers.

[astroman]

that's the overall latency if you feed audio through a VST processor.
You could do that with a Scope synth (for example) by feeding it's output to an Asio channel and monitor the Asio 'return' signal via Scope mixer - which would be a pretty dull setup
You'd rather monitor the synth directly with almost unmeasurable latency, and feed the direct out via Asio to the sequencer.
Depending on your overall setup some latency adjustment would be due.
If you trigger a VSTi synth/sampler you have only the output latency as delay between the Asio host/sequencer and Scope.
In those cases where you apply VST effects you have both input and output latency.

Personally I prefer to monitor (and time-adjust) everything in Scope and have the sequencer record 'behind' the already aligned sources.
In that case latency isn't a concern at all and Scope synths have the most direct feel (which is pointless with my abilities...)
But on guitar and bass I fully subscribe to your ideas about latency, as I'm used to cans in that case...
I really need some time to self adjust to the air-travel latency of a real amp. It just feels strange in the beginning.

cheers, Tom

[tgstgs]

NO_

my test is a simple *.exe file sending midievent from host to scope and back again measuring the time_

it is as fast as the host cpu is able to process the interups been called;

but in case of the vsti stuff talk with steinbegr;

it would be VERY very funny (i dont know!!!) if they collecting the midievents the same way they do with audiosamples_

but who knows . . .

good vibes

[tgstgs]

DELETED_vibes

[garyb]

Gary,

Last time I've checked, speed of sound in air is about 340 metres per second...


Marcos.

well, that's true. 1ms per foot. i wrote per second by mistake, but now it's corrected.

[CreamWare4Ever]

In those cases where you apply VST effects you have both input and output latency.

You mean to external signal, right? Adding VST effects to VSTi:s will not increase latency as default.

[CreamWare4Ever]

Personally I prefer to monitor (and time-adjust) everything in Scope and have the sequencer record 'behind' the already aligned sources.
In that case latency isn't a concern at all and Scope synths have the most direct feel (which is pointless with my abilities...)
But on guitar and bass I fully subscribe to your ideas about latency, as I'm used to cans in that case...
I really need some time to self adjust to the air-travel latency of a real amp. It just feels strange in the beginning.

This is my easy solution, and I think that I don't need to think about time-adjust at all:


Just 'V' connection straight into ASIO input